4 Empirical studies 


In this section the 5 studies the present dissertation is based on will be briefly summarized. While the first three studies aimed to contribute additional evidence regarding the neural network supporting vibrotactile WM, the goal of the last two studies was to investigate the role of the TOE and associated LTM representations in vibrotactile WM.

4.1 Neural correlates of human vibrotactile working

4.1.1 Study I: Neural correlates of vibrotactile working memory in the human brain

In the first study, fMRI was used to further investigate the neural network supporting vibrotactile WM in the human brain. Whereas single-unit recordings only allow studying neural activity in pre-selected brain regions such as S1 or lateral PFC, fMRI enables the investigation of the functioning of the whole brain in healthy human subjects. Therefore, fMRI provides the opportunity to identify additional regions that are part of the vibrotactile WM network. As the study by Klingberg and colleagues suggested, the PPC might also be involved in vibrotactile WM (Klingberg et al., 96). This suggestion is supported by studies using phonological and visuo-spatial WM in which association areas of the PPC have also been found to play a role (Smith und Jonides, 98). In addition, it has been shown that the recruitment of sensory brain regions in WM tasks depends on the length of the delay period. While ultra-short delay periods have been shown to be associated with activation of sensory brain regions, tasks with longer delay periods additionally recruit the lateral PFC (Cornette et al., 02; Pasternak und Greenlee, 05).

In this fMRI study, an event-related design was used to separate brain activity related to the different phases of the vibrotactile delayed discrimination task (encoding, maintenance, decision making). To identify the specific regions of the network crucially involved in WM functions, a WM condition was compared to a control condition which did not require maintaining the vibrotactile stimulus in WM. In addition, trials with short (100 ms) and long (4100 ms) delay periods were compared.


A distributed network of brain regions was associated with the performance of the vibrotactile delayed discrimination task. When activity in the WM condition was compared to activity in the control condition, only a few areas remained significantly activated. During encoding, consistent with neurophysiological data in monkeys and behavioral studies in humans, S1 and the ventral PMC exhibited increased activity. Because the ventral PMC (Broca's area) supports language processes in humans (Friederici, 02), its activation could indicate that the encoding of a vibrotactile stimulus in human subjects includes verbal processes. Maintenance of a vibrotactile memory trace evoked activity in PMC and lateral PFC. Decision making caused activation in S1, S2, PMC, and lateral PFC. However, human vibrotactile WM recruited additional areas. Decision making, in particular, activated a much broader network than known from the studies in non-human primates. Maintenance and decision making additionally activated the inferior PPC. The activation of the PPC is consistent with a previous study (Klingberg et al., 96) but now could be related to the maintenance and decision making periods of the task. This is in line with findings that link activity in the PCC to the active maintenance of information and decision making in other modalities (Gold und Shadlen, 07; Smith und Jonides, 98). Although the different task components evoked activity in distinctive neural networks, there was considerable overlap of activity, especially regarding maintenance and decision making, indicating that similar neural mechanisms are required for the sub-processes related to these task components.

Also, the comparison of short and long delay trials revealed specific activation patterns: While long delay trials had more anterior activation peaks in the lateral PFC and recruited the PPC, the opposite pattern was found for short delay trials. Short delay trials caused more posterior activation of the lateral PFC extending into premotor areas, while the parietal activation extended into S1. The finding that short delay trials are associated with increased levels of activity in S1 indicates that primary sensory regions are more crucial for performing a WM task with extremely short delay periods and confirms WM studies in the visual modality (Cornette et al., 02). However, due to the low temporal resolution of fMRI, this S1 activity can not be associated with a specific task period. Therefore, methods with a better temporal resolution than fMRI are required to investigate the role of S1 for the maintenance of the vibrotactile memory trace in the delay period.

Together, the fMRI study showed that a much broader network than previously studied supports the performance during the different periods of the vibrotactile delayed discrimination task. The finding that the PPC supports maintenance and decision making is of particular interest because posterior modality-specific association areas have also been implicated for sustained maintenance and decision making in other modalities (Pasternak und Greenlee, 05; Gold und Shadlen, 07). This indicates structural similarities in the neural organization of WM between modalities.

4.1.2 Study II: Rolandic and posterior-parietal alpha rhythms are differentially modulated during vibrotactile working memory 


As mentioned in the introduction, single-unit recordings in monkeys as well as behavioral and TMS studies in humans produced inconsistent results regarding the importance of S1 in the early delay period (Harris et al., 02; Romo und Salinas, 03). The fMRI study (Study I) revealed that S1 is crucially involved in the encoding of the standard stimulus and that S1 is more important for short WM trials as compared to long WM trials. However, due to the low temporal resolution of fMRI it is not possible to disentangle activity during the encoding of the standard stimulus and the early delay period. Therefore, in the second study the excellent temporal resolution of EEG was used to shed light on the role of S1 in the early delay period.

The rolandic alpha and beta rhythms over the somatomotor cortex indicate the physiological state of the underlying cortex with low power indicating active processing or readiness to process somatosensory stimuli and high power indicating low levels of activity (Hari et al., 97; Pfurtscheller et al., 97). In addition, high levels of alpha power over a brain region have been suggested to reflect the functional inhibition of this region during cognitive processing (Klimesch et al., 07). This functional inhibition could serve either to keep out new sensory information that could be further transmitted to areas actively maintaining the memory trace and thus interfering with the maintenance process or to shut-down irrelevant areas in order to devote processing capacity to areas actively involved in maintenance (Jokisch und Jensen, 07; Tuladhar et al., 07; Klimesch et al., 07).

In this study, EEG was recorded to investigate the dynamics of the rolandic rhythms during the encoding and maintenance phase of the vibrotactile delayed discrimination task. Based on the functional inhibition account, a reduction of rolandic rhythms over the somatomotor cortex during the delay period, representing maintenance of the vibrotactile memory trace, was hypothesized. A second question concerned the role of S1 during the early delay period. It was expected that if the neural representation of the vibrotactile stimulus is being maintained or still actively processed (consolidated) in S1 during the early delay, the rolandic rhythms, especially the alpha component, should still be suppressed over contralateral somatomotor cortex during this time period. To investigate if S1 activity is differentially modulated when subjects have to retain the memory trace for a short as opposed to longer delay, again two different delay lengths (370 ms and 4070 ms) were used.


Vibrotactile stimulation led to the typical modulation of the rolandic alpha and beta rhythms over contralateral somatomotor cortex with initial power suppression and a significant post-stimulus power increase which was stronger in the beta than in the alpha frequency band. In addition, during the middle portion of the delay period alpha power over the somatomotor cortex significantly increased over baseline levels. In light of the functional inhibition hypothesis, this could be interpreted as relative inactivation of the somatosensory cortex during memory retention. Across all subjects, the rolandic rhythms were not significantly modulated during the first second of the delay period. However, power was still reduced compared to baseline. Also, when compared to low performing subjects, higher performing subjects exhibited higher alpha amplitudes over contralateral somatomotor cortex during the late encoding period and the early delay period. In addition, high performers reached their alpha peak earlier than low performers. This suggests that the divergent findings in humans and monkeys might be caused by differences in encoding efficiency. Due to faster and more efficient encoding in high performing or trained subjects, consolidation of the neural code might already be finished during stimulus presentation and the memory trace is already maintained in other brain regions. The most pronounced effect for the rolandic alpha rhythm is a reduction of baseline power before the upcoming standard stimulus in the WM compared to a No-WM condition. This might reflect a tonic up-regulation of the activation level of contralateral S1 caused by sustained attention.

Concurrently, with the rolandic alpha suppression during stimulation, an alpha power increase was present at posterior parietal sites. This posterior parietal alpha power remained increased during the entire delay period indicating that posterior parietal and occipital visual areas are functionally inhibited during the active maintenance of a vibrotactile stimulus. In addition, the posterior alpha power increase was accompanied by a prefrontal alpha power increase. Higher prefrontal alpha power was also found during the encoding and early delay period of long as compared to short WM trials probably reflecting increased executive demands when a vibration frequency has to be maintained for a longer time. Therefore, simultaneous increases at prefrontal and posterior parietal sites might reflect top-down control by a fronto-parietal network (Palva und Palva, 07).

In summary, somatosensory and visual brain regions seem to be functionally inhibited during the delay period of the vibrotactile WM task. Concerning the question regarding the role of S1 during the early delay period, this study indicates that S1 is not involved in active maintenance of the vibrotactile stimulus. The activation level of S1 during the early delay period depends on the efficiency of stimulus encoding. Top-down attentional control probably up-regulates the activation level of S1 under WM demands in order to optimize task performance.

4.1.3 Study III: Pre-trial subliminal electrical stimulation improves performance in vibrotactile delayed discrimination 


The results of Harris and colleagues (02), who applied TMS to contralateral S1, indicate that S1 supports the maintenance of vibrotactile information during the early delay period. However, the results of the EEG study (Study II) indicated that S1 does not actively maintain the memory trace and that the level of activity in S1 during the early reflects ongoing consolidation processing in untrained or low performing subjects. The effects of TMS are not very localized and not specific concerning the type of neurons affected (Siebner und Rothwell, 03). In contrast, subliminal electrical stimulation leads to weak but selective inhibition of S1 (Taskin et al., 08; Blankenburg et al., 03). Therefore, it can be used to further investigate the proposed ongoing memory consolidation during the early delay period. In addition, Study II indicated that pre-trial activity of S1 plays an important role for task processing. However, the question remains whether a manipulation of the physiological state of S1 affects performance in the vibrotactile discrimination task.

To address these issues, subliminal stimulation was applied concurrently during the pre-trial, encoding and early delay period of the vibrotactile delayed discrimination task. To prove that the efficiency of a subliminal train of pulses lasting only 1 s could produce similar effects on detection-threshold as continuous stimulation (Blankenburg et al., 2003; Taskin et al., 08), an initial pilot experiment using a modified non-blocked version of the previous detection experiment was conducted. In the pilot experiment, subliminal stimulation raised the detection threshold for subsequent weak electrical stimuli confirming and extending the results of previous studies in which block-wise subliminal stimulation was used (Taskin et al., 08; Blankenburg et al., 03). In contrast, when subliminal stimulation was applied in the pre-trial period of a vibrotactile delayed discrimination task, performance improvements were observed. There is converging evidence that the detection of weak stimuli and the processing and encoding of strong stimuli are associated with different physiological brain states (Nicolelis, 05). It has been suggested that subliminal electrical stimulation reduces background noise in S1 (Blankenburg et al., 03). This facilitates the encoding of the upcoming strong vibratory stimulus. In the case of the weak electrical stimuli in the pilot experiment, the reduction of noise levels prevents mechanisms of stochastic resonance. Stochastic resonance has been shown to improve the detection of weak stimuli (Collins et al., 96; Collins et al., 95). When subliminal stimulation was applied during the early delay period, performance impairments were observed for a subgroup of subjects who received (defined by the slope of their psychometric function) the strongest subliminal stimulation. For this subgroup the results are in line with results by Harris and colleagues (02) who found that single pulses of TMS, when applied over contralateral S1 during the early delay period, disrupted performance in the vibrotactile delayed discrimination task. The effect of subliminal stimulation during the early delay period depends on the intensity of subliminal stimulation possibly reflecting the level of inhibition achieved. Therefore, the present study indicates that relatively strong subliminal stimulation interferes with the ongoing processing of a fragile vibrotactile stimulus representation in S1 during the early delay period.

In summary, study III indicates that the memory representation is still consolidated but not actively maintained in S1 during the early delay period confirming the results of the EEG study (Study II). Furthermore, this study confirms the finding of Study II regarding the importance of the physiological state of S1 in the pre-trial period. However, whereas in Study II top-down modulation enhances the activity in S1, here less noisy bottom-up processing facilitates performance.

4.2 Time order effects in vibrotactile working memory: the influence of long-term memory representations on task performance

4.2.1 Study IV: Time-order effect in vibrotactile delayed discrimination: The influence of trial types and response alternatives


The TOE phenomenon has been found consistently in psychophysical experiments in various modalities (Hellstrom, 85). In the tactile domain, the TOE has already been reported for vibrotactile discrimination (Sinclair und Burton, 96). Burton and Sinclair used a wide frequency range with standard frequencies of 50, 100, and 200 Hz including flutter and higher frequency vibrations. In addition, the difference between standard and comparison stimuli in their study was quite large ranging from 14 to 46 Hz difference which makes it more likely that subjects categorize the second stimulus as high or low instead of discriminating between the stimuli (Hernandez et al., 97). It remains an open question whether the TOE can be also found when only frequencies within the flutter range are used. In addition, it is not clear how changes in task parameters affect the TOE. In this series of behavioral experiments, the TOE was investigated in more detail. An initial experiment was carried out to determine whether the TOE can be found in vibrotactile delayed discrimination tasks when only stimuli in the range of flutter frequency (i.e., 10 - 50 Hz) are used. To prevent the subjects from using categorical judgments to solve the task, 6 standard frequencies and smaller frequency differences between standard and comparison stimulus ranging from 1 to 7 Hz were used. Three further experiments were conducted to investigate whether the TOE is affected by manipulations of the trial types and response alternatives available to the subject. Depending on the experiment, the trial types and response alternatives were symmetrical (e.g., "comparison stimulus is higher", "comparison stimulus is lower", and "both stimuli have the same vibration frequency") or asymmetrical (e.g., "comparison is higher", "both stimuli have the same vibration frequency").

Despite the variation in trial types and the available response alternatives, the TOE was consistently found in all 4 experiments. In addition, the magnitude of the TOE increased with increasing delay length and decreasing frequency difference between the two stimuli. Together, these results confirm that the TOE is a robust phenomenon in tasks involving the discrimination of magnitudes even within the range of flutter frequency. The independence of the TOE from the response alternatives and the pattern of results suggest that subjects implicitly use information about the average vibration frequency when performing the task. The question remains: In which period of the vibrotactile delayed discrimination task does average information affect task performance leading to the TOE? Whereas decisional accounts of the TOE assume that average information is used during the decision making process (Allan, 79; Masin und Fanton, 89), sensory accounts propose that the current standard vibration is integrated with the average vibration during encoding and maintenance (Hellstrom, 85; Helson, 64).

4.2.2 Study V: Neural correlates of time order effects: Implicit sensory stimulus representations influence vibrotactile delayed discrimination

This study aimed to identify the brain structures that are related to the generation of the TOE in the vibrotactile delayed discrimination task using fMRI. Perceptual accounts of the TOE suggest that it results from weighting processes during the encoding and maintenance of the standard stimulus (Hellstrom, 85). These weighting processes seem to reflect the integration of the sensory evidence provided by the standard stimulus and the weighted average of all stimuli presented previously. To identify brain regions that are involved in this assumed weighting process, the data from the first fMRI study (Study I) were re-analyzed. First, brain regions in which the BOLD response varies parametrically with the deviation of the vibration frequency of the standard stimulus from an assumed implicit average frequency were identified. To do this, the absolute value of the difference between the standard frequency and the arithmetic mean of the stimulus set was calculated with higher values reflecting larger deviations of the standard vibration from the mean vibration. The resulting values were then used as a parametric regressor in the subsequent fMRI analysis. The weighting processes leading to the TOE affect the difficulty of the subsequent decision. This is indicated by the finding that accuracy is determined by the time-order in which the stimuli are presented. To additionally identify regions showing a parametric modulation of the BOLD signal depending on the difficulty of the decision during the decision period, a second parametric regressor was used. The values of this parametric regressor were calculated by subtracting the comparison frequency from the mean frequency.


Brain activity in S1 and S2, the thalamus and the ventral PMC showed a parametric modulation during the encoding period, while the PPC showed a corresponding pattern during the maintenance period. Importantly, the BOLD signal in S2 during encoding and PPC during maintenance predicted individual differences in the size of the TOE providing the necessary link between the theoretical assumption of this integration process and actual behavior. The only region showing an effect of decision difficulty was the anterior cingulate cortex (ACC). The ACC has been associated with detection and monitoring of response conflicts, uncertainty and error detection (Braver et al., 01; Barch et al., 01; Ridderinkhof et al., 04; Botvinick et al., 99; Yoshida und Ishii, 06). Supporting this view, the reaction time pattern reveals that in addition to lower accuracy, subjects showed increased response times for decisions where the un-preferred time-order had to be judged. This suggests that subjects experience a conflict about their decision in trials when, in the case of the un-preferred time-order, information provided by the first vibration and the average information diverges.

Together, this study indicates that vibrotactile decision making is influenced by implicit LTM representations of the average vibration frequency. The neural network related to somatosensory processing seems to weight information about this average and the current standard frequency before the comparison period which supports perceptual accounts of the TOE. In addition, the present neural and behavioral data suggest that this conflict processing is an important factor accompanying the decision process in magnitude discrimination tasks in addition to earlier perception-related processes.

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 4.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML-Version erstellt am: